CN109300030A - Method and device for realizing stock investment recommendation - Google Patents

Abstract

The invention discloses a method and a device for realizing stock investment recommendation. The method includes: acquiring a given stock set; calculating a rise and fall probability for each stock in the stock set; and according to the rise and fall of each stock in the stock set Probability, select one or more stocks for investment advice, which is convenient, fast, and highly accurate, and can help investors to more accurately understand market trends and stock dynamics for investors or stock market analysts.

Description

Method and device for realizing stock investment recommendation

technical field

The present invention relates to the fields of artificial intelligence and big data, and in particular to a method, device, electronic device and computer-readable storage medium for realizing stock investment recommendation.

Background technique

Investors usually use search engines to find relevant value information to help their final decision-making, and most of these decision-making processes rely on human analytical judgment and experience. In fact, stock review data on the Internet contains rich and valuable semantic information that can help investors understand market trends and stock dynamics. Existing stock review analysis methods usually only focus on capturing the sentiment polarity of stock reviews, so as to understand the macro effect of stock reviews on market trends. However, stock reviews on the Internet often contain a lot of noise, such as water army and individual subjective tendency to conform to the herd, which seriously affects the judgment of investors. Therefore, it is very meaningful to use artificial intelligence technology to conduct fine-grained and authoritative analysis of stock review information, and then automatically select high-quality stocks from massive information for investors and stock analysts.

SUMMARY OF THE INVENTION

In view of the above problems, the present invention is proposed to provide a method, apparatus, electronic device and computer-readable storage medium for implementing stock investment recommendation that overcome the above problems or at least partially solve the above problems.

According to one aspect of the present invention, a method for implementing stock investment recommendation is provided, the method comprising:

Get the given stock collection;

Calculate the probability of ups and downs for each stock in the stock set;

According to the rising and falling probability of each stock in the stock set, one or more stocks are selected for investment advice.

According to another aspect of the present invention, there is provided a device for implementing stock investment recommendation, the device comprising:

A stock set acquisition unit, adapted to acquire a given stock set;

a rise and fall probability calculation unit, adapted to calculate the rise and fall probability for each stock in the stock set;

The stock investment recommendation unit is adapted to select one or more stocks for investment advice according to the rising and falling probability of each stock in the stock set.

According to yet another aspect of the present invention, there is provided an electronic device, the electronic device comprising: a processor, and a memory storing a computer program executable on the processor;

Wherein, the processor is configured to execute any one of the methods described above when executing the computer program in the memory.

According to yet another aspect of the present invention, there is provided a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements any of the methods described above.

According to the technical solution of the present invention, by obtaining a given stock set; calculating the probability of ups and downs for each stock in the stock set; selecting one or more stocks for investment according to the ups and downs of each stock in the stock set Advice is convenient, quick, and highly accurate, and can help investors more accurately understand market trends and stock dynamics for investors or stock market analysts.

The above description is only an overview of the technical solutions of the present invention, in order to be able to understand the technical means of the present invention more clearly, it can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand , the following specific embodiments of the present invention are given.

Description of drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are for the purpose of illustrating preferred embodiments only and are not to be considered limiting of the invention. Also, the same components are denoted by the same reference numerals throughout the drawings. In the attached image:

1 shows a flowchart of a method for implementing stock investment recommendation according to an embodiment of the present invention;

Figure 2 is a schematic diagram of a stock comment data information;

3 is a schematic diagram of another representation of a piece of stock review data information;

Figure 4 is a schematic diagram of the original stock review data volume and the cleaned stock review data volume;

Figure 5 is a schematic diagram of the profit situation after selecting stocks using the intelligent stock selection method c;

6 shows a schematic diagram of an apparatus for implementing stock investment recommendation according to an embodiment of the present invention;

FIG. 7 shows a schematic diagram of another device for implementing stock investment recommendation according to an embodiment of the present invention;

Fig. 8 shows a schematic diagram of yet another device for implementing stock investment recommendation according to an embodiment of the present invention;

9 is a schematic structural diagram of an electronic device in an embodiment of the present invention;

FIG. 10 is a schematic structural diagram of a computer-readable storage medium in an embodiment of the present invention.

Detailed ways

Explanation of terms appearing in the present invention:

FM: Factorization Machine, factorization machine, is a well-known algorithm, a machine learning algorithm based on matrix decomposition proposed by Steffen Rendle, which is widely used in classification and prediction models.

SVM: Support Vector Machine, support vector machine, is a well-known algorithm and a common discrimination method. In the field of machine learning, it is a supervised learning model that is usually used for pattern recognition, classification and regression analysis.

ARMA: Auto Regressive Moving Average, an autoregressive moving average model, one of the high-resolution spectral analysis methods of the model parameter method. This method is a typical method to study the rational spectrum of stationary stochastic processes, and is suitable for a large class of practical problems. Compared with AR model method and MA model method, it has more accurate spectral estimation and better spectral resolution performance.

OSRatio: Opinion Shift Ratio, the Opinion Change Ratio, is used to characterize the likelihood that a stock commentator will change his opinion on the same stock.

TSRatio: the Ratio of True-then-Shift, the ratio of correct opinions to change, is used to characterize the possibility of a stock commentator changing his opinion on the premise that the stock commentary's opinion is correct.

FSRatio: the Ratio of False-then-Shift, which is the ratio of changing wrong views, which is used to indicate the possibility of a stock commentator changing his views on the premise of a wrong view of a stock review.

TCTRatio: the Reliability Ratio of True-then-Constant, the reliability ratio of unanimously correct opinions, is used to characterize the reliability of a stock commentator's opinion on a stock comment under the premise that the opinion is correct.

TSTRatio: the Reliability Ratio of True-then-Shift, the reliability ratio of changing correct opinions, is used to characterize the reliability of changing opinions on the premise that the opinions of stock commentators are correct.

FCTRatio: the Reliability Ratio of False-then-Constant, the unanimous error reliability ratio, is used to characterize the reliability of the stock commentator's opinion on the stock comment under the premise of being wrong.

FSTRatio: the Reliability Ratio of False-then-Shift, the reliability ratio of changing wrong opinions, which is used to characterize the reliability of changing opinions under the premise that stock commentators are wrong about stock comments.

BIC Criterion: Bayesian Information Criterion, Bayesian Information Criterion. Bayesian decision theory is an important part of subjective Bayesian induction theory. It is to estimate the partially unknown state with subjective probability under incomplete intelligence, then use the Bayesian formula to modify the probability of occurrence, and finally use the expected value and the modified probability to make the optimal decision.

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be more thoroughly understood, and will fully convey the scope of the present disclosure to those skilled in the art.

FIG. 1 shows a flowchart of a method for implementing stock investment recommendation according to an embodiment of the present invention. As shown in FIG. 1 , the method includes:

Step S11: obtain a given stock set;

Step S12: Calculate the probability of ups and downs for each stock in the stock set;

Step S13: Select one or more stocks for investment advice according to the rising and falling probability of each stock in the stock set. This step includes:

According to the probability of rising and falling of each stock in the stock set, select one or more stocks for investment advice, including one or more of the following:

Select the preset number of stocks with the highest probability of rising and rising for investment advice, and choose the average weighting method for investment weight;

Select the preset number of stocks with the highest probability of rising and rising for investment advice, and the investment weight is selected according to the probability of rising;

Select a stock with the highest probability of rising and rising from each stock sector, and choose an average weighting method for investment weight;

Select a stock with the highest probability of rising and rising from each stock sector, and the investment weight will be weighted according to the probability of rising;

Select one or more stocks with the highest rising probability from each stock sector, choose an average weighting method among the sectors, and choose a weighting method according to the rising probability among the selected stocks in each sector.

In an embodiment of the present invention, selecting one or more stocks for investment advice includes one or more of the following according to the rising and falling probability of each stock in the stock set:

Get a dataset of stock reviews for the same stock s _j

Calculate the probability of the stock going up or down cf(s _j ) according to the following formula:

in, Represents the stock reviews dataset The number of stock review data in , c _i represents a piece of stock review data, Opinion polarity for the stock comment data, is the reliability index of the stock review data, _rυ (ci ) is the exact value of the reliability classification of the stock review data;

When cf(s _j )≥0, the stock s _j goes up, and the probability of going up is |cf(s _j )|;

When cf(s _j )<0, the stock s _j falls, and the probability of falling is |cf(s _j )|.

By obtaining a given stock set; calculating the probability of rising and falling for each stock in the stock set; selecting one or more stocks for investment advice according to the rising and falling probability of each stock in the stock set, which is convenient, fast and accurate High, which can help investors understand market trends and stock dynamics more accurately for investors or stock market analysts.

In one embodiment of the present invention, the embodiment shown in FIG. 1 determines the opinion polarity of a piece of stock review data according to the following method

Obtain a training set and a validation set consisting of stock review data, and label opinion polarity for each stock review data in the training set and validation set;

Based on the labeled training set, the machine learning model is trained, and the effect of the learning model is evaluated based on the labeled test set, and the trained machine learning model is obtained;

Input the relevant information of the stock review data to be predicted into the trained machine learning model, obtain the opinion polarity classification information of the stock review data output by the machine learning model, and determine the stock review data according to the opinion polarity classification information view polarity.

In an embodiment of the present invention, the reliability index of a piece of stock review data is determined according to the following formula

in, represents the date, Yes stock price, Yes the stock price the next day, is stock comment opinion polarity

In an embodiment of the present invention, the embodiment shown in FIG. 1 determines the exact value _rυ (ci ) for reliability classification of a piece of stock review data according to the following method:

Extract feature vector based on stock review dataset and stock price sequence set;

Use the extracted feature vector to train the support vector machine SVM model based on radial basis kernel function;

Train a machine learning model for predicting stock prices using stock price sequence sets;

Integrate the SVM model and the machine learning model for predicting stock prices to obtain a classification model for evaluating the reliability of stock reviews

but The larger the value of _rυ (ci ), the more reliable the classification result of the reliability of stock reviews.

In an embodiment of the present invention, in the embodiment shown in FIG. 1, extracting a feature vector based on a stock review data set and a stock price sequence set includes:

Based on each piece of stock review data in at least part of the stock review data set, one or more of the following features are extracted to form a feature vector:

The bullish or bearish view polarity information of the stock comment data;

Among all the stock comment data for stock s released on the day t, the number of bullish stock comment data and the number of bearish stock comment data;

The number of bullish stock reviews, the number of bearish stock reviews, the number of stock reviews with correct views, and the number of stock reviews with wrong views among all stock review data for stock s published within the first preset length of time from day t The number of stock review data;

The price sequence of the stock s in the second preset length of time since day t;

The price of the stock s predicted by the machine learning model used to predict the stock price on the next trading day and the standard deviation of the output of the model;

From the third preset length of time from day t, among all the stock comment data published by stock commentator a, the number of bullish stock comment data, the number of bearish stock comment data, the number of stock comment data with correct view, and the number of incorrect view data The number of stock review data;

From the fourth preset length of time since day t, among the stock comment data for stock s released by stock commentator a, the number of bullish stock comment data, the number of bearish stock comment data, and the number of correct stock comment data and the number of misinformed stock review data;

Determined based on the stock comment sequence published by stock commentator a in the fifth preset length of time from day t, based on the probability of change of opinion of stock commentator a, OSRatio, probability of change of opinion under the premise of correct opinion TSRatio, opinion Probability of changing your opinion under the wrong premise FSRatio, maintaining your opinion if you are right and keeping your opinion correct TCTRatio, changing your opinion if you are right and changing your opinion is correct TSTRatio, maintaining your opinion if you are wrong and maintaining your opinion One or more of the probability FCTRatio of maintaining the correct view and the probability FSTRatio of changing the view and changing the view if the view is wrong;

Among them, the stock commentator of the stock comment data is a, the comment is the stock s, and the release date is t.

In one embodiment of the present invention, in the embodiment shown in FIG. 1, using the extracted feature vector to train the SVM model based on the radial basis kernel function includes:

Let the radial basis kernel function be:

The SVM model is:

Among them, x ₁ and x ₂ are two feature vectors, and Y is the parameter of the radial basis kernel function; the function φ( ) maps the original feature to the high-dimensional kernel space for the calculation of the optimal decision hyperplane;

The parameters ω and b of the SVM model are calculated by optimizing the objective function as follows:

sry _i (ω ^T φ(ci _{)+b)≥l-ξ i ,}

ξ _i ≥ 0, i=1,...,N

where C is the trade-off parameter for noise versus simplified hyperplane classification in the training samples, and _yi is the label for whether the stock review opinion is correct.

In one embodiment of the present invention, in the embodiment shown in FIG. 1 , the machine learning model for predicting stock prices using stock price sequence sets to train includes:

Determine the stock price sequence data as the model training set and test set, wherein each piece of data in the training set or test set includes: the stock closing price of several consecutive days used to input the model, and the stock closing price of the next day as a label;

Train the ARMA model based on the training set and verify the prediction effect of the model based on the validation set

In one embodiment of the present invention, integrating the SVM model and the machine learning model for predicting stock prices, the obtained classification model for evaluating the reliability of stock reviews includes:

Based on the stock price prediction results of the machine learning model used to predict stock prices, the following classification equation is constructed:

in, Yes time stock price, is predicted by a machine learning model used to predict stock prices the stock price the next day, is the stock review opinion polarity, and err( _ci ) is the standard deviation of the current stock forecast price output by the machine learning model used to predict stock prices;

Integrate an SVM model and a machine learning model for predicting stock prices: where u∈[0,1];

The final classification model for evaluating the reliability of stock reviews is:

Among them, when h(c _i ) is 1, it indicates that the stock rating is reliable; when h( _ci ) is -1, it indicates that the stock rating is unreliable

In an embodiment of the present invention, the embodiment shown in FIG. 1 determines the bullish or bearish opinion polarity information of the piece of stock review data in the following manner:

Obtain a training set and a validation set consisting of stock review data, and label opinion polarity for each stock review data in the training set and validation set;

Based on the labeled training set, the machine learning model is trained, and the effect of the learning model is evaluated based on the labeled test set, and the trained machine learning model for predicting the opinion polarity of the stock review data is obtained;

Input the stock review data into the machine learning model for predicting the opinion polarity of the stock review data, and obtain the opinion polarity classification information of the stock review data output by the model

In an embodiment of the present invention, the embodiment shown in FIG. 1 determines the opinion polarity distribution information of stock commentator a based on the following method:

Based on each adjacent stock comment data in the stock comment sequence of the same stock by stock commentator a, the stock comment data pair is extracted;

Based on the extracted stock review data pairs, determine the probability of change of opinion of the stock commentator a, OSRatio, probability of change of opinion under the premise of correct opinion, probability of change of opinion under the premise of correct opinion, probability of change of opinion under the premise of wrong opinion FSRatio, probability of change of opinion under the premise of correct opinion, and maintenance of opinion under the premise of correct opinion. Probability of correct opinion TCTRatio, Probability of changing opinion under the premise of correct opinion and changing opinion is correct TSTRatio, Probability of maintaining opinion under the premise of wrong opinion and remaining correct opinion and Probability of changing opinion under the premise of wrong opinion and changing opinion Correct probability FSTRatio.

The solution for reliability modeling of stock review data proposed by the present invention is a unified framework that integrates a variety of heterogeneous information sources, such as stock price time series, stock review text content, and stock reviews that publish stock reviews. It can effectively filter noise and screen out valuable and reliable stock review information for investors or stock market analysts; it can not only be used for reliability analysis of stock review information, but also in other aspects of the financial field. Such as economic situation analysis, accurate stock recommendation, portfolio management and automatic trading. The specific implementation scheme is as follows:

1. Cleaning and processing of stock review data. Through data cleaning, the noise of stock review data obtained from the Internet can be initially cleaned, including:

(1) Delete stock review data with neutral opinion polarity.

(2) Delete the sequence data and stock review data corresponding to the stock review sequence whose length is less than 5.

Figure 2 is a schematic diagram of a stock review data information. As shown in Figure 2, a stock review text includes stock reviewer 201 (allan), time 202 (8 days ago), opinion polarity 203 (BUY, Bullish), target stock 204 (IBM ), comment content 205 (I think there is a support at 173.11) and other information.

Among them, when the opinion polarity is neutral, it is difficult to be automatically identified, that is, deleting the stock review data whose opinion polarity is neutral requires manual screening. "Stock review sequence with length less than 5" means that the number of reviews on the same stock by the same stock reviewer is less than 5.

Figure 3 is a schematic diagram of another representation of a stock review data information. It can be seen from the figure that the target stock is classified as A shares. 2016-12-29, the opinion polarity is bullish, including the content of opinion polarity: the stock price encounters the support of the annual line, you can consider buying, the opinion is for reference.

Figure 4 is a schematic diagram of the original stock review data volume and the cleaned stock review data volume, and the data source is the Sina Financial Planner website. As can be seen from the figure, the number of cleaned data is greatly reduced, which removes a large amount of noise in the stock review data, thereby reducing the amount of calculation for subsequent data processing.

2. Mining of stock commentator opinion polarity and reliability distribution pattern, the polarity tendency and reliability distribution of stock commentary can be mined through historical stock comment information of stock commentator, including:

(1) Statistics of the stock comment polarity distribution of the stock commentator by the stock commentator's historical stock comment information, that is, the release of the bullish and bearish probability distribution. Mining the opinion polarity distribution information of stock commentators includes one or more of four modes, which are briefly summarized as: one-to-one, one-to-many, many-to-one and many-to-many, specifically:

Based on all historical stock review data of the same stock reviewer for the same stock in the acquired stock review data, determine the probability that the stock reviewer will issue bullish stock review data for the stock, and determine that the stock reviewer will issue bullish stock review data for the stock Probability of bearish stock review data;

Based on all historical stock review data of the same stock reviewer for different stocks in the obtained stock review data, determine the probability that the stock reviewer will issue bullish stock review data, and determine the probability that the stock reviewer will issue bearish stock review data probability;

Based on all historical stock review data for the same stock by different stock reviewers in the obtained stock review data, determine the probability that a stock reviewer will issue bullish stock review data for the stock, and determine that a stock reviewer will issue a bearish stock review data for the stock Probability of stock review data;

Based on all historical stock review data for different stocks by different stock reviewers in the acquired stock review data, a probability of posting bullish stock review data and a probability of posting bearish stock review data are determined.

(2) Statistics of the reliability distribution of the stock reviews by the stock reviewer through the stock reviewer's historical stock review information, that is, the reliability and unreliable probability distribution of the stock reviews.

3. Mining of opinion consistency patterns of stock commentators, mining their opinion consistency probability distributions through historical stock review sequence data of stock commentators, including:

(1) Based on the adjacent stock comment data in the stock comment sequence of the same stock commentator by the same stock commentator, extract the stock comment data pair, that is, the 2-gram data pair, which is the stock comment data pair containing the opinion polarity. ;

(2) Based on the extracted stock review data pairs, calculate the probability that the stock reviewer maintains his opinion and the probability that he changes his opinion.

For example, the adjacent stock comment data in the stock comment sequence of the same stock commentator are: bullish, bearish, bearish, bullish, bullish. Based on the above data, the 2-gram data pairs of opinion polarities are obtained, which are respectively : bullish, bearish; bearish, bearish; bearish, bullish; bullish, bullish. Based on the above 2-gram data pair, the probability that the stock commentator maintains his or her opinion is calculated, that is, the probability of unanimous opinion is 0.5, and the probability of changing his opinion is 0.5.

4. Mining of opinion change patterns of stock commentators, mining their opinion change patterns through historical stock comment sequence data of stock commentators, including:

(1) Based on the adjacent stock comment data in the stock comment sequence of the same stock commentator for the same stock, extract the stock comment data pair, that is, use the comment sequence data of the stock commentator on the same stock to extract opinion polarity and correct opinion or not two 2-gram data pairs;

(2) Based on the extracted stock comment data pairs, determine the probability TSRatio of the stock commentator to change his opinion under the premise of being correct, and determine the probability FSRatio of the stock commentator to change his opinion under the premise of being wrong, that is, according to the extreme Sexual data pair statistics on the probability TSRatio of changing opinions under the premise of correct opinions, and the probability FSRatio of changing opinions under the premise of wrong opinions;

(3) Based on the extracted stock comment data pairs, determine the stock commentator maintains his opinion on the premise that his opinion is correct, and the probability TCTRatio of the maintained opinion is correct, and determine that the stock commentator changes his opinion on the premise that the opinion is correct, and The probability TSTRatio of the changed opinion is correct, that is, the reliability of maintaining the opinion based on the data on the premise that the statistical opinion is correct. reliability of opinion TSTRatio;

(4) Based on the extracted stock comment data pairs, determine the probability FCTRatio that the stock commentator maintains his opinion under the premise of being wrong, and the probability FCTRatio that the maintained opinion is correct, and determines that the stock commentator changes his opinion under the premise of being wrong, and The probability FSTRatio of the changed opinion is correct, that is, the reliability of maintaining the opinion under the premise that the statistical opinion is wrong according to the data FCTRatio (that is, the stock commentator's opinion was wrong at the previous moment, and the opinion is still correct at the next moment), and the opinion is changed under the premise of wrong opinion. Opinion reliability FSTRatio.

For example, the adjacent stock comment data in the stock comment sequence of the same stock commentator are: bullish, bearish, bearish, bullish, bullish. Based on the above data, the 2-gram data pairs of opinion polarities are obtained, which are respectively : bullish, bearish; bearish, bearish; bearish, bullish; bullish, bullish, at the same time get the 2-gram data pair of whether the opinion is correct or not, corresponding respectively: correct, correct; wrong, correct; correct, wrong; correct, correct.

According to the polarity of opinion data, the probability TSRatio of changing opinions on the premise of correct opinions is 0.5, and the probability FSRatio of changing opinions under the premise of wrong opinions is 0; the reliability of maintaining opinions on the premise that statistical opinions are correct according to the data TCTRatio is 0.25, opinions The reliability TSTRatio of changing opinions under the correct premise is 0.25; the reliability FCTRatio of maintaining opinions under the premise of incorrect statistical opinions according to the data is 0.25, and the reliability of changing opinions under the premise of incorrect opinions is 0.

Five, stock review opinion polarity detection (o(ci)), using the collected historical stock review text data to train the FM model, and based on the trained FM model to classify and predict the opinion polarity of the stock review data, where the FM model is the machine The learning model is an existing algorithm model, but it is specially processed in the present invention and applied to the polarity detection of stock opinions, which specifically includes:

(1) Obtain a training set and a validation set consisting of stock review texts, and annotate opinion polarity for each stock review text in the training set and validation set, that is, determine the training set, development set and test set of stock review texts, in which the development The set is similar to the test set and is collectively referred to as the validation set. Among them, the development set is used to optimize the model parameters during the training process to obtain the optimal model, and the test set is used to test the effect of the model after training; the labeling of opinion polarity is manual labeling, that is, the training set and the test set are manually labeled Concentrates the sentiment polarity (bullish or bearish) of each stock review text.

(2) Perform word segmentation on the text of the training set, and obtain a dictionary based on statistics. For example, "I think the stock will rise tomorrow". The separable words are: "I", "think", "tomorrow", "stock", "will", "Rising", similar to the word segmentation method, obtains a dictionary by statistics.

(3) Based on the dictionary, determine the TF-IDF feature of each stock review text in the training set, the feature is a vector of the size of the dictionary, and each dimension is the TF-IDF value of the corresponding word based on the text.

TF-IDF (term frequency–inverse document frequency) is a commonly used weighting technique for information retrieval and data mining. TF means Term Frequency, and IDF means Inverse Document Frequency. The main idea of TFIDF is: if a word or phrase appears frequently in one article and rarely appears in other articles, it is considered that the word or phrase has a good ability to distinguish between categories and is suitable for classification. TFIDF is actually: TF*IDF, TF Term Frequency, IDF Inverse Document Frequency. TF represents the frequency of the term appearing in document d. The main idea of IDF is: if there are fewer documents containing the term t, that is, the smaller the n, the larger the IDF, it means that the term t has a good ability to distinguish between categories. If the number of documents containing term t in a certain type of document C is m, and the total number of documents containing t in other types is k, obviously the number of documents containing t is n=m+k. When m is large, n is also large. , the value of the IDF obtained according to the IDF formula will be small, which means that the classification ability of the entry t is not strong.

To put it simply, common words that appear frequently in some other documents in the training set, such as "的", "le", etc., are of low importance, while "bullish", "bearish" appearing in stock review texts. Opinion polar words such as ” are more important. TF-IDF is a feature that evaluates the importance of each word in the dictionary.

Regarding the TF-IDF feature is a vector of the size of the dictionary, each dimension is the understanding of the corresponding word based on the TF-IDF value of the text. For example, there are 1000 words in 100 sentences, then the vector of each sentence It is 1000-dimensional, for example, the initial vector is [1, 0, 0, ... 1], where 1 means that the target word appears in the sentence, 0 means that the target word does not appear in the sentence, and both 1 and 0 in the initial vector must be Multiply the TF-IDF value of the stock review text, that is, multiply the weight of the word, to get the TF-IDF feature of the stock review text.

(4) Extract features from the stock review text of the training set, use the extracted features as the input of the machine learning model, and use the opinion polarity classification information of the stock review text as the output of the machine learning model; about the TF of the stock review text in the training set - The IDF feature is used as the model input feature, and the sentiment polarity of the stock comment is the output, that is, the output is bullish or bearish, that is, the output is 1 or 0.

(5) Based on the opinion polarity classification information of the output of the machine learning model and the opinion polarity marked by the corresponding stock review text, calculate the loss of the machine learning model, and learn the parameters of the machine learning model based on the calculated loss; that is, based on the training set , using the adaptive regularization stochastic gradient descent method to learn the FM model parameters, and using cross-validation to optimize and adjust the value of the hyperparameter k in the FM model, where the value of the hyperparameter k is a manually given value.

(6) Based on the validation set, evaluate the effect of the FM model, specifically: extracting features from the stock review text of the validation set, inputting the extracted features into the machine learning model, and obtaining the opinion polarity of the stock review text output by the machine learning model Classification information: Based on the opinion polarity classification information output by the machine learning model and the opinion polarity marked by the corresponding stock review text, evaluate the effect of the machine learning model.

(7) Repeat (5), (6) and (7) until the FM effect meets the requirements (eg, the accuracy rate is greater than 95%), then the FM model training is completed.

(8) Based on the trained FM model, the opinion polarity classification is performed on the stock review text, and the o(ci) attribute is obtained.

(9) Calculate the reliability r(ci) of each stock review according to (Equation 1):

in, represents the date, for stock price, for the stock price the next day, is 0 or 1.

(10) Generate corresponding structured data for the stock review text, the structured data includes: stock reviewer identification, review time, target stock, opinion polarity and reliability index, that is, construct stock review unit data ci={d( ci),a(ci),s(ci),t(ci),o(ci),r(ci)}, where d(ci) is the comment content, a(ci) is the stock commentator identifier, s (ci) is the target stock, t(ci) is the comment time, o(ci) is the opinion polarity, and r(ci) is the reliability index.

6. The method of scoring the reliability of stock review information, that is, to score the reliability of a certain stock review information of a certain stock reviewer. Extract key features from stock rating sequence, stock price sequence and stock reviewer historical behavior data, and score the reliability of stock rating information based on the integrated learning framework of classification model and time series analysis model, including:

(1) Extract feature vectors based on the stock review data set and the stock price sequence set. First, based on each piece of stock review data in at least part of the stock review data in the stock review data set, extract one or more of the following features to form a feature vector:

The bullish or bearish opinion polarity information of the stock review data; how to determine the bullish or bearish opinion polarity information of the stock review data has been elaborated in step 5, and will not be repeated here.

Among all the stock comment data for stock s released on the day t, the number of bullish stock comment data and the number of bearish stock comment data;

The number of bullish stock reviews, the number of bearish stock reviews, the number of stock reviews with correct views, and the number of stock reviews with wrong views among all stock review data for stock s published within the first preset length of time from day t The number of stock review data;

The price sequence of the stock s in the second preset length of time since day t;

The price of the stock s predicted by the machine learning model used to predict the stock price on the next trading day and the standard deviation of the output of the model;

From the third preset length of time from day t, among all the stock comment data published by stock commentator a, the number of bullish stock comment data, the number of bearish stock comment data, the number of stock comment data with correct view, and the number of incorrect view data The number of stock review data;

From the fourth preset length of time since day t, among the stock comment data for stock s released by stock commentator a, the number of bullish stock comment data, the number of bearish stock comment data, and the number of correct stock comment data and the number of misinformed stock review data;

Determined based on the stock comment sequence published by stock commentator a in the fifth preset length of time from day t, based on the probability of change of opinion of stock commentator a, OSRatio, probability of change of opinion under the premise of correct opinion TSRatio, opinion Probability of changing your opinion under the wrong premise FSRatio, maintaining your opinion if you are right and keeping your opinion correct TCTRatio, changing your opinion if you are right and changing your opinion is correct TSTRatio, maintaining your opinion if you are wrong and maintaining your opinion One or more of the probability FCTRatio of maintaining the correct view and the probability FSTRatio of changing the view and changing the view if the view is wrong;

Among them, the stock commentator of the stock comment data is a, the comment is the stock s, and the release date is t.

How to determine the opinion polarity distribution information of stock commentator a has been explained in detail in step 3, and will not be repeated here.

For example, key features are extracted from stock review sequence, stock price sequence and stock commentator historical behavior data, the key features include: opinion polarity, historical stock status, price time series and stock commentator historical behavior. Among them, the opinion polarity is bullish or bearish of the current comment; the historical stock status includes two situations: the first is that regardless of time, among all the stock comment data for stock s released on that day, the number of bullish stock comment data, the number of bearish ones The number of stock review data; the second is the number of bullish stock review data, the number of bearish stock review data, the number of stock review data with correct views, and the number of stock review data with correct views and wrong views among all stock review data for stock s in the past 7 days. The number of stock comment data; the price time series includes: the price series of the stock s in the past 25 days and the standard deviation of the next day's price and output predicted by the ARMA model; the historical behavior of stock commentators includes: a stock commentator a in the past 7/ Number of bullish/bearish/correct/incorrect stock reviews made in 30/90 days; number of bullish/bearish/correct/incorrect stock reviews made by a stock analyst on the current stock in the past 7/30/90 days; based on a stock One or more of OSRatio, TSRatio, FSRatio, TCTRatio, TSTRatio as determined by commentator a's sequence of stock reviews posted in the past 7/30/90 days.

(2) Use the extracted feature vector to train the support vector machine SVM model based on the radial basis kernel function (Equation 2):

Let the radial basis kernel function be:

Among them, x ₁ and x ₂ are two feature vectors, which can also become variables; Y is the parameter of the radial basis kernel function, which is generally set to 1 divided by the total number of features, such as 10,000 features, then r is set to 0.0001; φ( ) maps the original features to the high-dimensional kernel space, so as to facilitate the calculation of the optimal decision hyperplane (equation 3);

The SVM model is:

The principle of SVM is to solve the separating hyperplane that can correctly divide the training dataset and has the largest geometric interval. The input is some feature sample points, the model is learning a hyperplane, this hyperplane can determine two things: 1. All data points are perfectly divided into two categories, the output of the first category is 1 (corresponding to reliable stock reviews), the first The output of the second category is 0 (corresponding to unreliable stock reviews). 2. The farther away all data points are from the hyperplane, the better.

If the feature sample point is linearly inseparable in the original space (in most cases, it is linearly inseparable), then we hope to map it to the high-dimensional space through a mapping to make the problem linearly separable, using The mapping of is the kernel function.

(3) Calculate the parameters ω and b by optimizing (Equation 4):

sty _i (ω ^T φ(ci )+b)≥1-ξ _{i ,}

ξ _i ≥0, i=1,...,N, (Equation 4)

where C is the trade-off parameter for noise vs. simplified hyperplane classification in the training samples, and _yi is the label for whether the stock review opinion is correct. The three parameters ω, b, and ξ are all parameters that need to be learned by model training, where ω and b are the two parameters used by the SVM model in prediction; st represents the previous constraints, that is, the latter two parameters. Rows are the constraints of the objective function in the first row. y _i is the boundary of the objective function, and the bigger the boundary, the better.

(4) Use stock price sequence sets to train machine learning models for predicting stock prices, such as ARMA models, including:

a. Determine the stock price sequence data of the training set and the test set, the input data is a number of consecutive stock closing prices, and the output is the stock closing price of the next day; that is, determine the stock price sequence data as the model training set and test set, among which the training set or Each piece of data in the test set includes: the stock closing price of several consecutive days used to input the model, and the stock closing price of the next day as a label;

b. Train the ARMA model based on the training set, and verify the prediction effect of the model based on the validation set; that is, based on the training set, use the maximum likelihood estimation to train the ARMA model parameters, and adjust the parameters p and q based on the BIC criterion. The ARMA model uses the historical stock price data of a stock to predict the stock price of the next day, and verifies the prediction effect based on the validation set.

In general, the stock price prediction based on the time series analysis model uses the historical price series of the stock to train the ARMA model, and predicts the price of the stock one day later based on the trained ARMA model.

(5) Integrate the SVM model and the machine learning model for predicting stock prices to obtain a classification model for evaluating the reliability of stock reviews; that is, construct a classification equation based on the stock price prediction results, as shown in Equation 5 below:

in, Yes time stock price, Yes The predicted value of the stock price for the next day, is the sentiment polarity of stock reviews, and err( _ci ) is the standard deviation of the stock price sequence data, that is, the error or confidence value of the stock price prediction value currently output by the model.

(6) Integrate the SVM model and the ARMA model to obtain the final classification function, as shown in Equation 6 below:

When h(c _i ) is 1, it indicates that the stock rating is reliable; when h( _ci ) is -1, it indicates that the stock rating is unreliable. in The calculation formula is as follows:

In formula 7, u∈[0, 1] is the weighting coefficient of the prediction results of the SVM and ARMA models, and it is determined by experiments that u=0.59 has the best effect.

The accurate value of the stock rating reliability classification can be calculated according to the following formula 8:

The higher the _rυ (ci ), the more reliable the results of rating the stocks. (Equation 8) is the absolute value of the output result of (Equation 7).

7. Calculation of the probability of a stock going up or down, calculating the probability of a stock going up or down through the relevant features and measurement results extracted in the process of measuring the reliability of the stock rating, including:

(1) Calculate the rising and falling probability cf(s _j ) of the stock according to the following formula 9:

in, Represents the stock reviews dataset The number of stock review data in , that is, the sum of all stock reviews, c _i represents a stock review data, Opinion polarity for the stock comment data, is the reliability index of the stock review data, and _rυ (ci ) is the exact value of the reliability classification of the stock review data.

(2) Predict the rise and fall of the stock according to the following formula 10:

(3) Calculate the probability of the stock going up or down according to Equation 11 below:

w(s _j )=|cf(s _j )|.(Equation 11)

When cf(s _j ) ≥ 0, the larger the value of w(s _j ), the greater the probability of the stock going up; when cf(s _j ) < 0, the greater the value of w(s _j ), indicating that the stock The probability of falling is high.

8. After the stock review reliability model is completed, when a query request for information about a stock reviewer's designated opinion is received, the result data corresponding to the query request can be output.

9. Stock investment based on the reliability model of stock rating, screening reliable stock rating based on the reliability model of stock rating data, and investing accordingly, including:

(1) Calculate the probability w(s _j ) of the stock rising or falling for all stocks in the stock pool, where s _j is a single stock;

(2) A variety of intelligent stock selection methods:

a. Select the preset number of stocks with the highest probability of rising and rising for investment advice, and choose the average weighting method for investment weight; that is, select the K stocks with the highest rising index as investment recommendations, and select the average weighting for investment weight. method, that is, the average investment in each stock is G/K yuan, where G is the total investment amount;

b. Select the preset number of stocks with the highest probability of rising and rising for investment advice, and the investment weight is selected in a way that is weighted according to the probability of rising; that is, the K stocks with the highest rising index are selected as investment recommendations, and the investment weight is selected According to the weighted method of increase index, that is, stock s _j investment Yuan

c. Select a stock with the highest probability of rising and rising from each stock sector, and choose an average weighting method for investment weight; that is, select a stock with the highest growth index in each sector as an investment recommendation, with a total of M ( M = 10) sectors (see Table 1 below), and the investment weights are equally weighted, that is, each stock invests G/M yuan.

Table 1: Sectors of stock symbols

Table 1 is the stock section information, Category represents the section name, and #Covered Symbols represents the number of stocks in the section.

d. Select a stock with the highest probability of rising and rising from each stock sector, and the investment weight will be weighted according to the probability of rising; that is, select a stock with the highest rising index in each sector as an investment suggestion, There are a total of M (M=10) sectors, and the investment weights are chosen to be equally weighted, that is, each stock s _j invests in

e. Select one or more stocks with the highest probability of rising and rising from each stock sector, choose an average weighting method among the sectors, and select stocks in each sector that are weighted according to the probability of rising. method; that is, a combination of the above stock selection methods, for example, first select Km stocks with the highest increase from each section, and then use the average weighting or weighting according to the increase index to invest in each stock. The total investment in each section can also be weighted equally or weighted by the growth index.

Figure 5 is a schematic diagram of the profit situation after using the intelligent stock selection method c to select stocks. From January 2016 to December 2016, the intelligent stock selection method c was selected for simulated investment, and K stocks were selected for investment on each trading day. The profit situation is as follows As shown in Figure 5, the total investment is 10,000 yuan, K=M, and each stock is 10,000/M.

FIG. 6 shows a schematic diagram of an apparatus for implementing stock investment recommendation according to an embodiment of the present invention, wherein the apparatus 600 includes:

A stock set acquisition unit 601, adapted to acquire a given stock set;

A rise and fall probability calculation unit 602, adapted to calculate a rise and fall probability for each stock in the stock set;

The stock investment recommendation unit 603 is adapted to select one or more stocks for investment advice according to the rising and falling probability of each stock in the stock set.

In an embodiment of the present invention, the stock investment recommendation unit is adapted to perform stock investment recommendation in one or more of the following ways:

Select the preset number of stocks with the highest probability of rising and rising for investment advice, and choose the average weighting method for investment weight;

Select the preset number of stocks with the highest probability of rising and rising for investment advice, and the investment weight is selected according to the probability of rising;

Select a stock with the highest probability of rising and rising from each stock sector, and choose an average weighting method for investment weight;

Select a stock with the highest probability of rising and rising from each stock sector, and the investment weight will be weighted according to the probability of rising;

Select one or more stocks with the highest rising probability from each stock sector, choose an average weighting method among the sectors, and choose a weighting method according to the rising probability among the selected stocks in each sector.

7 shows a schematic diagram of another apparatus for implementing stock investment recommendation according to an embodiment of the present invention. The apparatus 70 includes: a stock set acquisition unit 601 , a rise and fall probability calculation unit 602 and a stock investment recommendation unit 603 . Wherein, the rise and fall probability calculation unit 602 includes:

The obtaining unit 701 is adapted to obtain a stock comment data set for the same stock s _j

The calculation unit 702 is adapted to calculate the rising and falling probability cf(s _j ) of the stock according to the following formula:

in, Represents the stock reviews dataset The number of stock review data in , c _i represents a piece of stock review data, Opinion polarity for the stock comment data, is the reliability index of the stock review data, _rυ (ci ) is the exact value of the reliability classification of the stock review data;

When cf(s _j )≥0, the stock s _j goes up, and the probability of going up is |cf(s _j )|;

When cf(s _j )<0, the stock s _j falls, and the probability of rising is |cf(s _j )|.

8 shows a schematic diagram of another apparatus for implementing stock investment recommendation according to an embodiment of the present invention. The apparatus 80 includes: a stock set acquisition unit 601 , a rise and fall probability calculation unit 602 and a stock investment recommendation unit 603 . The rise and fall probability calculation unit 602 includes: an acquisition unit 701 and a calculation unit 702 , wherein the calculation unit 702 includes an opinion polarity prediction unit 801 , a stock rating reliability determination unit 802 , and a stock rating reliability classification unit 803 .

The opinion polarity prediction unit 801 is adapted to determine the opinion polarity of a piece of stock review data according to the following method

Obtain a training set and a validation set consisting of stock review data, and label opinion polarity for each stock review data in the training set and validation set;

Based on the labeled training set, the machine learning model is trained, and the effect of the learning model is evaluated based on the labeled test set, and the trained machine learning model is obtained;

Input the relevant information of the stock review data to be predicted into the trained machine learning model, obtain the opinion polarity classification information of the stock review data output by the machine learning model, and determine the stock review data according to the opinion polarity classification information polarity of opinion

The stock review reliability determination unit 802 is adapted to determine the reliability index of a piece of stock review data according to the following formula

in, represents the date, Yes stock price, Yes the stock price the next day, is a stock commentary opinion.

The stock review reliability classification unit 803 is adapted to determine the exact value _rυ (ci ) for reliability classification of a piece of stock review data according to the following method:

Extract feature vector based on stock review dataset and stock price sequence set;

Use the extracted feature vector to train the support vector machine SVM model based on radial basis kernel function;

Train a machine learning model for predicting stock prices using stock price sequence sets;

Integrate the SVM model and the machine learning model for predicting stock prices to obtain a classification model for evaluating the reliability of stock reviews

but The larger the value of _rυ (ci ), the more reliable the classification result of the reliability of stock reviews.

The stock review reliability classification unit 803 is adapted to extract one or more of the following features to form a feature vector based on each piece of stock review data in at least part of the stock review data in the stock review data set:

The bullish or bearish view polarity information of the stock comment data;

Among all the stock comment data for stock s released on the day t, the number of bullish stock comment data and the number of bearish stock comment data;

The number of bullish stock reviews, the number of bearish stock reviews, the number of stock reviews with correct views, and the number of stock reviews with wrong views among all stock review data for stock s published within the first preset length of time from day t The number of stock review data;

The price sequence of the stock s in the second preset length of time since day t;

The price of the stock s predicted by the machine learning model used to predict the stock price on the next trading day and the standard deviation of the output of the model;

From the third preset length of time from day t, among all the stock comment data published by stock commentator a, the number of bullish stock comment data, the number of bearish stock comment data, the number of stock comment data with correct view, and the number of incorrect view data The number of stock review data;

From the fourth preset length of time since day t, among the stock comment data for stock s released by stock commentator a, the number of bullish stock comment data, the number of bearish stock comment data, and the number of correct stock comment data and the number of misinformed stock review data;

Determined based on the stock comment sequence published by stock commentator a in the fifth preset length of time from day t, based on the probability of change of opinion of stock commentator a, OSRatio, probability of change of opinion under the premise of correct opinion TSRatio, opinion Probability of changing your opinion under the wrong premise FSRatio, maintaining your opinion if you are right and keeping your opinion correct TCTRatio, changing your opinion if you are right and changing your opinion is correct TSTRatio, maintaining your opinion if you are wrong and maintaining your opinion One or more of the probability FCTRatio of maintaining the correct view and the probability FSTRatio of changing the view and changing the view if the view is wrong;

Among them, the stock commentator of the stock comment data is a, the comment is the stock s, and the release date is t.

The stock evaluation reliability classification unit 803 uses the extracted feature vector to train the SVM model based on the radial basis kernel function, specifically:

Let the radial basis kernel function be:

The SVM model is:

Among them, x ₁ and x ₂ are two eigenvectors, and γ is the parameter of the radial basis kernel function; the function φ( ) maps the original feature to the high-dimensional kernel space for the calculation of the optimal decision hyperplane;

The parameters ω and b of the SVM model are calculated by optimizing the objective function as follows:

sty _i (ω ^T φ(ci )+b)≥1-ξ _{i ,}

ξ _i ≥ 0, i=1,...,N,

where C is the trade-off parameter for noise versus simplified hyperplane classification in the training samples, and _yi is the label for whether the stock review opinion is correct.

The stock evaluation reliability classification unit 803 uses the stock price sequence set to train the machine learning model for predicting stock prices as follows:

Determine the stock price sequence data as the model training set and test set, wherein each piece of data in the training set or test set includes: the stock closing price of several consecutive days used to input the model, and the stock closing price of the next day as a label;

Train the ARMA model based on the training set and verify the prediction effect of the model based on the validation set

The stock review reliability classification unit 803 integrates the SVM model and the machine learning model for predicting stock prices, and obtains a classification model for evaluating the reliability of stock reviews as follows:

Based on the stock price prediction results of the machine learning model used to predict stock prices, the following classification equation is constructed:

in, Yes time stock price, is predicted by a machine learning model used to predict stock prices the stock price the next day, is the stock review opinion polarity, err( _ci ) is the standard deviation of the current stock forecast price output by the machine learning model used to predict stock prices;

Integrate an SVM model and a machine learning model for predicting stock prices: where u∈[0,1];

The final classification model for evaluating the reliability of stock reviews is:

Among them, when h(c _i ) is 1, it indicates that the stock rating is reliable; when h( _ci ) is -1, it indicates that the stock rating is unreliable

The stock review reliability classification unit 803 is adapted to determine the bullish or bearish opinion polarity information of the piece of stock review data in the following manner:

Obtain a training set and a validation set consisting of stock review data, and label opinion polarity for each stock review data in the training set and validation set;

Based on the labeled training set, the machine learning model is trained, and the effect of the learning model is evaluated based on the labeled test set, and the trained machine learning model for predicting the opinion polarity of the stock review data is obtained;

The stock review data is input into a machine learning model for predicting the opinion polarity of the stock review data, and the opinion polarity classification information of the stock review data output by the model is obtained.

The stock review reliability classification unit 803 is adapted to determine the opinion polarity distribution information of stock reviewer a based on the following method:

Based on each adjacent stock comment data in the stock comment sequence of the same stock by stock commentator a, the stock comment data pair is extracted;

Based on the extracted stock review data pairs, determine the probability of change of opinion of the stock commentator a, OSRatio, probability of change of opinion under the premise of correct opinion, probability of change of opinion under the premise of correct opinion, probability of change of opinion under the premise of wrong opinion FSRatio, probability of change of opinion under the premise of correct opinion, and maintenance of opinion under the premise of correct opinion. Probability of correct opinion TCTRatio, Probability of changing opinion under the premise of correct opinion and changing opinion is correct TSTRatio, Probability of maintaining opinion under the premise of wrong opinion and remaining correct opinion and Probability of changing opinion under the premise of wrong opinion and changing opinion Correct probability FSTRatio.

To sum up, by obtaining a given stock set; calculating the probability of ups and downs for each stock in the stock set; and selecting one or more stocks for investment advice according to the ups and downs of each stock in the stock set. The present invention performs special processing and training on the existing machine learning model, and performs opinion polarity classification prediction on the stock review data, so that the relevant information of the stock review text to be predicted can be input into the trained machine learning model, and then the machine can be obtained. The opinion polarity classification information of the stock review text output by the learning model, calculate the probability of stock price rise and fall, and make stock investment according to the probability of stock price change, which is convenient, fast, and highly accurate, and integrates a variety of heterogeneous information sources, such as stocks The price sequence, the text content of stock reviews, and the historical behavior of stock reviewers who published stock reviews, based on the multi-source heterogeneous big data, with the help of data mining technology to deeply analyze and extract key features, and use these features to measure the reliability of stock reviews. Effectively filter noise, screen out valuable and reliable stock review information from massive information, and select high-quality stocks, which can help investors more accurately understand market trends and stock dynamics for investors or stock market analysts. This method can be applied not only to the reliability analysis of stock review information, but also to other aspects of the financial field, such as economic situation analysis, accurate stock recommendation, portfolio management and automatic trading.

It should be noted:

The algorithms and displays provided herein are not inherently related to any particular computer, virtual appliance, or other device. Various general-purpose devices can also be used with the teachings based on this. The structure required to construct such a device is apparent from the above description. Furthermore, the present invention is not directed to any particular programming language. It should be understood that various programming languages may be used to implement the inventions described herein, and that the descriptions of specific languages above are intended to disclose the best mode for carrying out the invention.

In the description provided herein, numerous specific details are set forth. It will be understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

Similarly, it is to be understood that in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together into a single embodiment, figure, or its description. This disclosure, however, should not be construed as reflecting an intention that the invention as claimed requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the Detailed Description are hereby expressly incorporated into this Detailed Description, with each claim standing on its own as a separate embodiment of this invention.

Those skilled in the art will understand that the modules in the device in the embodiment can be adaptively changed and arranged in one or more devices different from the embodiment. The modules or units or components in the embodiments may be combined into one module or unit or component, and further they may be divided into multiple sub-modules or sub-units or sub-assemblies. All features disclosed in this specification (including accompanying claims, abstract and drawings) and any method so disclosed may be employed in any combination, unless at least some of such features and/or procedures or elements are mutually exclusive. All processes or units of equipment are combined. Each feature disclosed in this specification (including accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will appreciate that although some of the embodiments described herein include certain features, but not others, included in other embodiments, that combinations of features of different embodiments are intended to be within the scope of the invention within and form different embodiments. For example, in the following claims, any of the claimed embodiments may be used in any combination.

Various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) may be used in practice to implement some of the apparatus for photographing and recording text content, the electronic device, and the computer-readable storage medium according to the embodiments of the present invention Or some or all of the functionality of all components. The present invention can also be implemented as apparatus or apparatus programs (eg, computer programs and computer program products) for performing part or all of the methods described herein. Such a program implementing the present invention may be stored on a computer-readable medium, or may be in the form of one or more signals. Such signals may be downloaded from Internet sites, or provided on carrier signals, or in any other form.

For example, FIG. 9 is a schematic structural diagram of an electronic device in an embodiment of the present invention. The electronic device 900 includes a processor 910 , and a memory 920 storing a computer program executable on the processor 910 . The processor 910 is configured to execute each step of the method in the present invention when executing the computer program in the memory 920 . The memory 920 may be electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read Only Memory), EPROM, hard disk, or ROM. The memory 920 has storage space 930 for storing a computer program 931 for performing any of the method steps in the above-described methods. The computer program 931 can be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards or floppy disks. Such a computer program product is typically a computer-readable storage medium as described in FIG. 10 .

FIG. 10 is a schematic structural diagram of a computer-readable storage medium in an embodiment of the present invention. The computer-readable storage medium 1000 stores a computer program 931 for carrying out the method steps according to the invention, which can be read by the processor 910 of the electronic device 900 and, when executed by the electronic device 900, causes the electronic device 900 to To execute each step in the above-described method, specifically, the computing program 931 stored in the computer-readable storage medium can execute the method shown in any of the above-described embodiments. The computer program 931 may be compressed in a suitable form.

It should be noted that the above-described embodiments illustrate rather than limit the invention, and that alternative embodiments may be devised by those skilled in the art without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, and third, etc. do not denote any order. These words can be interpreted as names.

Claims (10)

1. A method for implementing stock investment recommendation, wherein the method comprises: Get the given stock collection; Calculate the probability of ups and downs for each stock in the stock set; According to the rising and falling probability of each stock in the stock set, one or more stocks are selected for investment advice. 2. The method according to claim 1, wherein, according to the rising and falling probability of each stock in the stock set, selecting one or more stocks for investment advice includes one or more of the following: Select the preset number of stocks with the highest probability of rising and rising for investment advice, and choose the average weighting method for investment weight; Select the preset number of stocks with the highest probability of rising and rising for investment advice, and the investment weight is selected according to the probability of rising; Select a stock with the highest probability of rising and rising from each stock sector, and choose an average weighting method for investment weight; Select a stock with the highest probability of rising and rising from each stock sector, and the investment weight will be weighted according to the probability of rising; Select one or more stocks with the highest rising probability from each stock sector, choose an average weighting method among the sectors, and choose a weighting method according to the rising probability among the selected stocks in each sector. 3. The method of claim 1, wherein the calculating a probability of ups and downs for each stock in the stock set comprises: Get a dataset of stock reviews for the same stock s _j Calculate the probability of the stock going up or down cf(s _j ) according to the following formula: in, Represents the stock reviews dataset The number of stock review data in , _si represents a piece of stock review data, Opinion polarity for the stock comment data, is the reliability index of the stock review data, _rυ (ci ) is the exact value of the reliability classification of the stock review data; When cf(s _j )≥0, the stock s _j goes up, and the probability of going up is |cf(s _j )|; When cf(s _j )<0, the stock s _j falls, and the probability of falling is |cf(s _j )|. 4. The method of claim 3, wherein the opinion polarity of a piece of stock review data is determined according to the following method Obtain a training set and a validation set consisting of stock review data, and label opinion polarity for each stock review data in the training set and validation set; Based on the marked training set, the machine learning model is trained, and the effect of the learning model is evaluated based on the marked test set to obtain the trained machine learning model; Input the relevant information of the stock review data to be predicted into the trained machine learning model, obtain the opinion polarity classification information of the stock review data output by the machine learning model, and determine the stock according to the opinion polarity classification information Opinion polarity of the review data. 5. The method of claim 3, wherein the reliability index of a piece of stock review data is determined according to the following formula in, represents the date, Yes stock price, Yes the stock price the next day, It is the polarity of stock comments. 6. The method of claim 3, wherein the exact value _rυ (ci) for reliability classification of a piece of stock review data is determined as follows: Extract feature vector based on stock review dataset and stock price sequence set; Use the extracted feature vector to train the support vector machine SVM model based on radial basis kernel function; Train a machine learning model for predicting stock prices using stock price sequence sets; Integrate the SVM model and the machine learning model for predicting stock prices to obtain a classification model for evaluating the reliability of stock reviews but The larger the value of _rυ (ci ), the more reliable the classification result of the reliability of stock reviews. 7. The method of claim 6, wherein the extracting feature vectors based on the stock review data set and the stock price sequence set comprises: Based on each piece of stock review data in at least part of the stock review data in the stock review data set, one or more of the following features are extracted to form a feature vector: The bullish or bearish view polarity information of the stock comment data; Among all the stock comment data for stock s released on the day t, the number of bullish stock comment data and the number of bearish stock comment data; The number of bullish stock reviews, the number of bearish stock reviews, the number of stock reviews with correct views, and the number of stock reviews with wrong views among all stock review data for stock s published within the first preset length of time from day t The number of stock review data; The price sequence of the stock s in the second preset length of time since day t; The price of the stock s predicted by the machine learning model for predicting stock prices on the next trading day and the standard deviation of the model output; From the third preset length of time from day t, among all the stock comment data published by stock commentator a, the number of bullish stock comment data, the number of bearish stock comment data, the number of stock comment data with correct view, and the number of incorrect view data The number of stock review data; From the fourth preset length of time since day t, among the stock comment data for stock s released by stock commentator a, the number of bullish stock comment data, the number of bearish stock comment data, and the number of correct stock comment data and the number of misinformed stock review data; Determined based on the stock comment sequence published by stock commentator a in the fifth preset length of time from day t, based on the probability of change of opinion of stock commentator a, OSRatio, probability of change of opinion under the premise of correct opinion TSRatio, opinion Probability of changing your opinion under the wrong premise FSRatio, maintaining your opinion if you are right and keeping your opinion correct TCTRatio, changing your opinion if you are right and changing your opinion is correct TSTRatio, maintaining your opinion if you are wrong and maintaining your opinion One or more of the probability FCTRatio of maintaining the correct view and the probability FSTRatio of changing the view and changing the view if the view is wrong; Among them, the stock commentator of the stock comment data is a, the comment is the stock s, and the release date is t. 8. A device for implementing stock investment recommendation, wherein the device comprises: A stock set acquisition unit, adapted to acquire a given stock set; a rise and fall probability calculation unit, adapted to calculate the rise and fall probability for each stock in the stock set; The stock investment recommendation unit is adapted to select one or more stocks for investment advice according to the rising and falling probability of each stock in the stock set. 9. An electronic device, characterized in that the electronic device comprises: a processor, and a memory storing a computer program executable on the processor; Wherein, the processor is configured to execute the method of any one of claims 1-7 when executing the computer program in the memory. 10. A computer-readable storage medium on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the method of any one of claims 1-7 is implemented.